Previous: Bias Frames
Up: WHAT EXTRA OBSERVATIONS ARE REQUIRED?
Next: Dark counts
Previous Page: Bias Frames
Next Page: Dark counts
As described in section 1 these are essential to calibrate the gains of the individual pixels, and in so doing, many if not most of the apparent defects in raw CCD frames can be laundered away. They need doing for each filter, and good flat field exposures should have a mean level of between 20,000 and 40,000 counts/pixel. Although flat fields (i.e. pixel gains) appear very stable (they have been used from run to run), it is recommended that a complete set of flat fields (i.e. at least one per filter) is obtained each night. This is best done from the twilight sky when exposures of at least one second can be used to give the required levels. For the broad-band set of filters this happens during a short window of about 15 minutes duration in evening and morning twilight, close to sunrise. For the narrow-band emission line interference filters this window is much larger with some filters giving acceptable count rates (i.e. 32,000 per second) well before sunset. The GLANCE command is used to check the levels through the various filters. This obtains and displays the frame without writing it to disk/tape - (see II.3.1).
Adam:> GLANCE 2.0 (sec,e.g.)
Try setting DIPARM 2 0 65535 with the default LUT and use DISP to re-display - you can tell from the colour if the level is correct. To be more quantitative, use the PHOTOM routine with the STATS option:
Adam:> PHOTOM (select ``STATS" with the joystick box switches; see II.4.3)
then save the frame with the KEEP command:
Adam:> KEEP (see II.5.4)
Proceed through the filter set in this way (reddest first at the end of the night, last at the beginning). If the flat fields appear statistically similar each night, they can be added together to produce master flat-fields of very high s/n. It is worth comparing beginning to end of night flat-fields to check for problems.
An alternative method is to use an automated procedure written by C D Pike called TWILIGHT. This does a flat field for up to 6 filters in a filter wheel and continues until, in evening twilight for instance, the counts in one filter drop below the saturation level of 65535 counts. This filter is observed again and the rate at which the intensity is decreasing is calculated. It has been shown that this rate applies to the usual broad band filters (UBVRIZ) and so the UT at which each may be observed at a good S/N is calculated. From experience so far the procedure should be started at about sunset and at 15 minutes before sunrise. It is done in the following way
Adam:> SET LOG/NEW
and answer the prompts for required filters, upper and lower acceptable count rates and maximum integration time., The lowest integration time is set to 5 seconds. Details of the run numbers, filters, exposure times and mean count level are printed in the Adam log which can be listed on the line printer at the end of the procedure by typing the following:
Adam:> SHOW LOG
To combine or average flat fields the following commands can be used:
Adam:> DEFINE OFF COMBINE/WORKSPACE = 50
Then answer prompts. Do NOT try and handle more than 5 files at a time. When you have finished, end the process with:
Adam:> FORGET/DICT OFF
The sky flat fields are preferable because of colour matching - the twilight sky bears a much closer resemblance to the night sky than does a tungsten lamp illuminating the dome interior. In addition the sky flat fields will remove most self- fringing. (The present RCA chip shows some, even with broad- band filters.) But dome flat fields are very convenient in that they can be done during the day. They can serve as backup to the sky flat fields - in general they remove pixel-to- pixel variations very effectively, but because of poor colour matching, particularly at the extreme blue and red ends, large scale gain variations remain. Dome flat fields can further provide a quick check that the system is operating normally and that gain and noise are expected. That are thus a recommended afternoon/early evening occupation. To carry them out, illuminate evenly (but dimly) a clear patch of dome in front of the telescope using a single free-standing tungsten lamp. (Note that the flourescent lighting is far too bright and has lines in its spectrum whilst floor lighting is too faint and does not have a high enough colour temperature. Use the GLANCE technique described above unless you know from the previous day just what exposures are right.